The explanation of "region-concurrency"'s effect on speed for TiDB Lightning (Part 2 §2.2.1) is misleading

[kennytm]

tidb-in-action/session2/chapter2/lightning-internal.md

Line 80 in 832572f

* 举例来说，若一次编码处理耗时 50 毫秒，那么每秒只能进行 20 次编码。若 `block-size` 为 64 KB，则单一 CPU 核每秒最多完成 1.28 MB 数据的编码处理。当 `region-concurrency` 设置为 60，则整体编码处理的极限速度约为每秒 75 MB。

Lightning's encoder operates on rows, not read-blocks. So mentioning "block-size" here is misleading people into thinking increasing read-block-size improves encoding performance, but in fact the factor cancels out with the hidden factor in "一次编码处理"

Better eliminate the block-size and make it like

• 举例来说，若编码一行处理耗时 0.2 毫秒，那么每秒只能编码 1 ÷ 0.2ms = 5000 行。若一行大小平均为 250 字节，则单一 CPU 核每秒最多完成 250 × 5000 = 1.25 MB 数据的编码处理。当 region-concurrency 设置为 60，则整体编码处理的极限速度约为每秒 1.25 MB × 60 = 75 MB。